Photonic neural network on silicon chip based on tunable filter and modulation method thereof

Abstract

The invention discloses a photonic neural network on a silicon chip based on a tunable filter, and a modulation method and device thereof, and the photonic neural network on the silicon chip comprises an MZI network structure, a photoelectric detector, a light source, and the tunable filter. The input end of the MZI network is a light wave input end, and the output end is connected with the input end of the photoelectric detector for performing linear operation on input light waves; the output end of the photoelectric detector is connected with the electrode input end of the tunable filter, and the photoelectric detector is used for converting an optical signal output after linear operation of the MZI network into an electric signal and providing a modulation signal for the tunable filter; the light source is arranged at an optical signal input end of the tunable filter and is used for providing an optical signal for the tunable filter; the tunable filter is used for modulating according to the electric signals and optical signals provided by the light source, and the output end outputs light waves. The photonic neural network is low in power consumption, low in time delay, high in energy efficiency ratio and high in precision.

Description

Technical field [0001] The present invention relates to the field of optical communication technologies, and more particularly to a photonic neural network on a silicon substrate based on a tunable filter and a modulation method thereof. Background technique [0002] In recent years, artificial intelligence and large data technology have developed rapidly, and higher requirements have been made to the computational power of the chip, especially the rate and power consumption. The traditional electronic chip is limited to Moore Law, and its integration and computational power are slow. The photonic chip, due to its ultra-high transmission speed, ultra-low power consumption, has received more and more attention. Among them, the silicon-based photonic chip is made based on the SOI, Silicon-on-Insulator material, compatible with the traditional CMOS (Complementary Metal-Oxide-Semiconductor) process, high-rate, low power, low power consumption , High integration and anti-electromagnet...

AI Technical Summary

Problems solved by technology

However, the nonlinear activation function part is still implemented by electrical methods, and it cannot be integrated with the matrix budget part on-chip, and there are still problems of low operation speed and high power consumption.

Images